Figure 5. Effects of sample size on ACR and dip tests of unimodality. A, population distribution for D and G, with equal sample sizes and standard deviations between the modes. B, population distribution for E and H, with equal standard deviations between the modes but the sample sizes are unequal, with a ratio of 2:3. C, population distribution for F and I, with equal samples sizes between the modes but the standard deviations are unequal, with a ratio of 2:1. D–F, boxplots of P-values from dip test of unimodality based on 1000 random samples per sample size form population distributions in A–C, respectively. G–I, boxplots of P-values from ACR test of unimodality based on 1000 random samples per sample size form population distributions in A–C, respectively. Samples drawn from the bimodal distributions tend to be falsely judged unimodal (i.e. P-values above 0.05) by the tests at small sample sizes. Red lines are for a P-value of 0.05.